Michael Shi Email and Phone Number

• Designed and implemented an object-detection analysis system for text and image data using Python and OpenCV.• Built a computer vision pipeline for segmentation, labeling, and detection of 3-dimensional shapes combining a fine-tunedMask Regional CNN model with PyTorch Vision and CLIP segmentation model provided by HuggingFace• Designed a high-fidelity data infrastructure to index images and objects detected within images using PostgreSQL and Weaviate databases to support clustering via DBSCAN and t-SNE and visual analysis w/ Plotly.js.

• Implemented automated tests including UI, Contract, and Integration tests in TypeScript with GraphQL for core features to validate brokerage accounts, transactions, and other private user information.• Developed and deployed an automated alerts system for end of day trading with Humio to alert on-call engineers for new version releases.

• Created a machine learning research experiment design using Reinforcement Learning and synthetic data generation for Credit Card Fraud detection.• Generated a deep learning model which had 95 percent labeling accuracy on real and synthetic credit card transaction data provided by Capital One using Keras, Pandas, NumPy, and OpenAI Gym libraries in Python.

• Researched and implemented genetic algorithms for discrete and continuous geospatial data in Java to solve optimization problems for scientific applications.• Reduced the latency for baseline genetic algorithms by 25% using the Nelder-Mead optimization method